Font Size: a A A

Research And Implementation Of Remote Sensing Image Target Recognition Technology For Port Intelligence Analysis

Posted on:2021-05-19Degree:MasterType:Thesis
Country:ChinaCandidate:L T JiangFull Text:PDF
GTID:2392330647957215Subject:Calculation software and theory
Abstract/Summary:PDF Full Text Request
In the complicated surrounding environment,it's essential to get Comprehensive military intelligence and make real-time military interpretation for the winning of wars.As an important combat tool,counting the number of ships and monitoring the ports can help predict the intent of operations,locating the ship can help to hit the targets precisely.Therefore,it's of great strategic significance to study the detection and identification of ship.Remote sensing image is an important mean of information acquisition,widely used in intelligence reconnaissance.This paper focuses on the detection and identification based on the analysis of port information in remote sensing images.The main work of this paper includes the following three aspects:Firstly,separating the sea and land using ensemble learning and semantic segmentation.It's difficult to extract ships near port because of the similar to features between ships and porte integrate multiple semantic segmentation models to separate sea and land and eliminate false detection targets on the land.In this way,we can reduce false alarm rates and land areas in large-scale remote sensing images,reduce detection areas and improve detection efficiency.Secondly,we proposes an end-to-end ship detection method to solve the slow detection speed of algorithm based on the candidate.The method in this paper can quickly extract targets by converts the object detection into regression problem without the candidate frames.It's difficult to detect small ship due to the multiple scales.This paper introduces a feature pyramid network to improves the detection effect of small targets,and enhances the robustness of scale changes.Finally,we make research on ship identification technology to identify targets quickly in a rapidly changing battlefield.we use binary neural networks for target classification tasks in this paper.Comparing to traditional neural networks,binary neural networks have greatly improved on speed.Based on the above research,this paper designs and implements a prototype of the port intelligence analysis system.We test the Prototype System,the experimental results show that the system can process large-scale remote sensing images in real time,and achieve efficient,accurate,and rapid detection of ship targets.
Keywords/Search Tags:Remote sensing image, sea-land segmentation, ship detection, feature pyramid, binary neural network
PDF Full Text Request
Related items